AI2D

AI2D - Benchmark Leaderboard & Model Performance | AI Stats

Models Using This Benchmark

Organisation	Model	Reported	Top Score	Info	Self Reported	Source
Qwen	Qwen 3.6 Plus	01 Apr 2026	94.40%	-	Yes	Source
OpenAI	GPT 4o Audio (2024-10-01)	01 Oct 2024	94.20%	inferred modality/version alias from gpt-4o-2024-08-06	Yes	Source
OpenAI	GPT 4o Search Preview	11 Mar 2025	94.20%	inferred modality/version alias from gpt-4o-2024-08-06	Yes	Source
OpenAI	GPT 4o Transcribe Diarize	15 Oct 2025	94.20%	inferred modality/version alias from gpt-4o-2024-08-06	Yes	Source
OpenAI	GPT 4o Transcribe	20 Mar 2025	94.20%	inferred modality/version alias from gpt-4o-2024-08-06	Yes	Source
OpenAI	GPT 4o (2024-08-06)	06 Aug 2024	94.20%	-	Yes	Source
OpenAI	GPT 4o Realtime Preview (2024-10-01)	01 Oct 2024	94.20%	inferred modality/version alias from gpt-4o-2024-08-06	Yes	Source
OpenAI	GPT 4o Audio (2025-06-03)	03 Jun 2025	94.20%	inferred modality/version alias from gpt-4o-2024-08-06	Yes	Source
OpenAI	GPT 4o Realtime Preview (2025-06-03)	03 Jun 2025	94.20%	inferred modality/version alias from gpt-4o-2024-08-06	Yes	Source
OpenAI	GPT 4o Audio (2024-12-17)	17 Dec 2024	94.20%	inferred modality/version alias from gpt-4o-2024-08-06	Yes	Source
Mistral	Pixtral Large	18 Nov 2024	93.80%	-	Yes	Source
Qwen	Qwen 3.5 122B A10B	24 Feb 2026	93.30%	-	Yes	Source
Mistral	Mistral Small 3.2	20 Jun 2025	92.91%	-	Yes	Source
Qwen	Qwen 3.5 27B	24 Feb 2026	92.90%	-	Yes	Source
Qwen	Qwen 3.5 Flash	23 Feb 2026	92.90%	inferred family alias from qwen3.5-27b (score=0.4147; benches=81)	Yes	Source
Qwen	Qwen 3.5 35B A3B	24 Feb 2026	92.60%	-	Yes	Source
Meta	Llama 3.2 90B Vision Instruct	-	92.30%	inferred high-confidence family alias from llama-3.2-90b-instruct (score=0.5873; benches=13)	Yes	Source
Meta	Llama 3.2 11B Vision Instruct	-	91.10%	inferred high-confidence family alias from llama-3.2-11b-instruct (score=0.5873; benches=11)	Yes	Source
Meta	Llama 3.2 1B Instruct	25 Sept 2024	91.10%	inferred family alias from llama-3.2-11b-instruct (score=0.4200; benches=11)	Yes	Source
Qwen	Qwen 3 VL 235B A22B Instruct	-	89.70%	-	Yes	-
Qwen	Qwen 3 VL 32B Instruct	-	89.50%	-	Yes	-
Qwen	Qwen 3 VL 235B A22B Thinking	-	89.20%	-	Yes	-
Qwen	Qwen 3 VL 32B Thinking	-	88.90%	-	Yes	-
Qwen	Qwen 2.5 Math 72B	-	88.40%	inferred high-confidence family alias from qwen2.5-vl-72b (score=0.4700; benches=30)	Yes	Source
Qwen	Qwen 2.5 Math RM 72B	-	88.40%	inferred family alias from qwen2.5-vl-72b (score=0.4092; benches=30)	Yes	Source
Qwen	Qwen 2.5 Math PRM 72B	-	88.40%	inferred family alias from qwen2.5-vl-72b (score=0.4050; benches=30)	Yes	Source
Qwen	Qwen 2.5 VL 72B Instruct	-	88.40%	-	Yes	Source
xAI	Grok 1.5V	12 Apr 2024	88.30%	-	Yes	Source
Qwen	Qwen 3 VL 30B A3B Thinking	-	86.90%	-	Yes	-
Qwen	Qwen 3 VL 8B Instruct	-	85.70%	-	Yes	-
Qwen	Qwen 3 VL 30B A3B Instruct	-	85%	-	Yes	-
Qwen	Qwen 3 VL Embedding 8B	-	84.90%	inferred high-confidence family alias from qwen3-vl-8b-thinking (score=0.5232; benches=50)	Yes	-
Qwen	Qwen 3 Guard Stream 4B	-	84.90%	inferred family alias from qwen3-vl-4b-thinking (score=0.3371; benches=48)	Yes	-
Qwen	Qwen 3 8B	-	84.90%	inferred high-confidence family alias from qwen3-vl-8b-thinking (score=0.4600; benches=50)	Yes	-
Qwen	Qwen 3 Guard Gen 4B	-	84.90%	inferred family alias from qwen3-vl-4b-thinking (score=0.3400; benches=48)	Yes	-
Qwen	Qwen 3 VL 4B Thinking	-	84.90%	-	Yes	-
Qwen	Qwen 3 4B SafeRL	-	84.90%	inferred family alias from qwen3-vl-4b-thinking (score=0.3850; benches=48)	Yes	-
Qwen	Qwen 3 Reranker 4B	-	84.90%	inferred family alias from qwen3-vl-4b-thinking (score=0.3850; benches=48)	Yes	-
Qwen	Qwen 3 Embedding 4B	-	84.90%	inferred family alias from qwen3-vl-4b-thinking (score=0.3850; benches=48)	Yes	-
Qwen	Qwen 3 Reranker 8B	-	84.90%	inferred family alias from qwen3-vl-8b-thinking (score=0.3850; benches=50)	Yes	-
Qwen	Qwen 3 Guard Gen 8B	-	84.90%	inferred family alias from qwen3-vl-8b-thinking (score=0.3400; benches=50)	Yes	-
Qwen	Qwen 3 4B Thinking 2507	-	84.90%	inferred family alias from qwen3-vl-4b-thinking (score=0.3462; benches=48)	Yes	-
Qwen	Qwen 3 Guard Stream 8B	-	84.90%	inferred family alias from qwen3-vl-8b-thinking (score=0.3371; benches=50)	Yes	-
Qwen	Qwen 3 Embedding 8B	-	84.90%	inferred family alias from qwen3-vl-8b-thinking (score=0.3850; benches=50)	Yes	-
Qwen	Qwen 3 VL 8B Thinking	-	84.90%	-	Yes	-
Qwen	Qwen 3 VL Reranker 8B	-	84.90%	inferred high-confidence family alias from qwen3-vl-8b-thinking (score=0.5275; benches=50)	Yes	-
Qwen	Qwen 3 4B Instruct 2507	-	84.90%	inferred family alias from qwen3-vl-4b-thinking (score=0.3462; benches=48)	Yes	-
Qwen	Qwen 3 VL 4B Instruct	-	84.10%	-	Yes	-
Qwen	Qwen 2.5 Coder 3B	-	83.20%	inferred family alias from qwen2.5-omni-7b (score=0.3000; benches=45)	Yes	Source
Qwen	Qwen 2.5 Math 7B	-	83.20%	inferred high-confidence family alias from qwen2.5-omni-7b (score=0.4767; benches=45)	Yes	Source
Qwen	Qwen 2.5 Omni 7B	-	83.20%	-	Yes	Source
Qwen	Qwen 2.5 Math PRM 7B	-	83.20%	inferred family alias from qwen2.5-omni-7b (score=0.4092; benches=45)	Yes	Source
Qwen	Qwen 2.5 Omni 3B	-	83.20%	inferred high-confidence family alias from qwen2.5-omni-7b (score=0.4933; benches=45)	Yes	Source
Qwen	Qwen 2.5 Math 7B PRM800K	-	83.20%	inferred family alias from qwen2.5-omni-7b (score=0.3696; benches=45)	Yes	Source
Qwen	Qwen 2.5 Coder 7B	-	83.20%	inferred high-confidence family alias from qwen2.5-omni-7b (score=0.4700; benches=45)	Yes	Source
Microsoft	Phi 4 multimodal instruct	01 Feb 2025	82.30%	-	Yes	Source
DeepSeek	DeepSeek VL2	13 Dec 2024	81.40%	-	Yes	Source
DeepSeek	DeepSeek VL2 Small	13 Dec 2024	80%	-	Yes	Source
Microsoft	Phi 3.5 vision instruct	23 Aug 2024	78.10%	-	Yes	Source
Microsoft	Phi 3 Vision 128K Instruct	-	78.10%	inferred family alias from phi-3.5-vision-instruct (score=0.3495; benches=9)	Yes	Source
DeepSeek	DeepSeek VL2 Tiny	13 Dec 2024	71.60%	-	Yes	Source

Recorded Results

Average Score

Score Range

Leading Model

Models Using This Benchmark